Spatial-Proteomics-Framework

A framework for processing spatial proteomics data from images to analysis. Modified from the ImcSegmentationPipeline from Bodenmiller Group.

Installation

1. Download CellProfiler

   • configure by selecting Files -> Preferences, set CellProfiler plugins directory to ImcPluginsCP/plugins and restart CellProfiler.

2. Download Ilastik

3. Run conda install imageio numpy pandas readimc scipy tifffile xtiff

Process 0: Pre-process experiment directory

1. Input Directory

   • Create a directory with some DIRECTORY_NAME, e.g. MB0002_1_345
   • Inside the directory, we require two files:
     1. Image file of either:
     • mcd file: avoid spaces in naming. This is the output from the Hyperion Imaging System
     • tiff file: avoid spaces in naming. This is the an alternative output.
     2. Panel file: must be named panel.csv. This should have four columns like as the following matrix.

        • Metal tag is the identifier on the antibody.

        • Target indicates a protein measured.

        • In the full column, the markers of interest are denoted with 1, others are 0.

        • In the ilastik column, the markers used for segmentation are denoted with 1, others are 0.

          Metal Tag	Target	full	ilastik
          Dy162	CD3	1	0
          Ir193	DNA	0	1
          Pt196	Membrane	0	1
          ...	...	...	...

2. Analysis Directory

   • In lieu of setting up an analysis directory manually, we automate the creation here. Depending on your image file type, use the following commands in your terminal:
   • mcd file: python 0_pre_processing.py -d DIRECTORY_NAME
   • tiff file: python 0B_pre_processing.py -d DIRECTORY_NAME

3. Output

   • All files will be in a directory named DIRECTORY_NAME_analysis
   • Inside the analysis directory, there should be five folders: cpinp, cpout, crops, ilastik, and ometiff.

Process 1: Train pixel classifier

1. CellProfiler
   • Open 1_prepare_ilastik.cppipe
   • Drop the directory DIRECTORY_NAME_analysis/ilastik into the window under Images
   • In Output Settings: Set input to DIRECTORY_NAME_analysis/cpinp
   • In Output Settings: Set output to DIRECTORY_NAME_analysis/crops
   • Press Analyze Images
2. Ilastik
   • Open the Ilastik software.
   • Create a Pixel Classification project. Put it in your analysis directory and name it as you wish.
   • Press Input Data -> Add New... -> Add separate Image(s) and select all .h5 files in analysis/crops.
   • Press Feature Selection and select all features with σ >= 1
   • Press Training:
     • Rename labels: "Label 1 - Nucleus", "Label 2 - Cytoplasm", "Label 3 - Background".
     • Scroll down to Raw Input in the Group Visibility box, press up and down to check the different Ilastik channels indicated in panel.csv
     • Label the pixels by pressing the label and then the brush icon, switch back and forth between channels under Raw Input, you can also erase labels using the eraser icon
     • Press Live Update to update segmentation
     • Turn on/off segmentation and prediction masks displayed on the image to check your work by pressing the eye icon in each category in the Group Visibility box
   • Press Prediction Export and configure as the following:
     • Source: Probabilities
     • Choose Export Image Settings: Convert to Data Type: unsigned 16-bit, check Renormalize, Format: tiff
     • Press export all.
   • Press Batch Processing and press Raw Data Files... and select all _s2.h5 files in analysis/ilastik then click on Process all files

Process 2 Cell segmentation

1. CellProfiler

   • Open 2_segment_ilastik.cppipe
   • Drop the directory DIRECTORY_NAME_analysis/ilastik into the window under Images
   • In Output Settings: Set input to DIRECTORY_NAME_analysis/cpinp
   • In Output Settings: Set output to DIRECTORY_NAME_analysis/cpout
   • Press Analyze Images

2. Output

   • All files will be in a directory named DIRECTORY_NAME_analysis/cpout
   • Masks and probabilities are in cpout/masks/ and cpout/probabilities/

Process 3: Cell quantification

1. CellProfiler

   • Open 3_measure_mask.cppipe
   • IMPORTANT: If you are using a tiff file OR you are receiving errors in image matching rules:
     • In Metadata: press No on the question Extract metadata?
     • In NamesAndTypes: set Image set matching method to Order instead of Metadata
   • Drop the directory DIRECTORY_NAME_analysis/cpout into the window under Images
   • In Output Settings: Set input to DIRECTORY_NAME_analysis/cpinp
   • In Output Settings: Set output to DIRECTORY_NAME_analysis/cpout
   • Set channels in the first MeasureObjectIntensityMultichannel to the number of full channels in panel.csv (default is 40)
   • Set channels in the first MeasureImageIntensityMultichannel to the number of full channels in panel.csv (default is 40)
   • Press Analyze Images

2. Output

   • All files will be in a directory named DIRECTORY_NAME_analysis/cpout
   • Expression features are stored in cpout/cell.csv
   • Spatial features are stored in cpout/Object relationships.csv